Topology and Network Resources Discovery Protocol for Content-Aware Networks Vlad Poenaru, Eugen Borcoci, Marius Zotea University POLITEHNICA of Bucharest
Authors affiliation: Vlad Poenaru, Ph.D student Eugen Borcoci, prof. Marius Zotea student Telecommunication Dept., Electronics Telecommunications and Information Technology Faculty, University Politehnica of Bucharest, Romania Vlad.Poenaru@elcom.pub.ro, Eugen.Borcoci@elcom.pub.ro Acknowledgment: This work has been partially supported by the European Research Integrated Project FP7 ALICANTE MediA Ecosystem Deployment Through Ubiquitous Content-Aware Network Environments 2010-2013 and partially by the national Romanian project POSDRU/89/76909 2
Main objectives The paper proposes and develops: a protocol for topology and network resources discovery in a multi-domain media oriented distribution eco-system QoS enabled spanning multiple IP domains starting from a previously defined architecture, the protocol is specified protocol design, implementation and some performance evaluation shortly presented 3
CONTENTS 1. ALICANTE project architecture (short- high level description) 2. TNRD Protocol Specification, Design and Implementation 3. Conclusions 4
1. ALICANTE System Architecture ALICANTE : New concepts (Future Internet oriented) Content Aware Networking (CAN) Network Aware Application (NAA) Novel virtual CAN layer (Data Plane + Mgmt&Ctrl Plane) on top of IP networks focused on multimedia distribution but not limited to, Quality of Services (QoS) assurance with different levels of guarantees In the Data Plane: Create Virtual Content Aware Networks (VCAN), multidomain, unicast/ multicast and QoS enabled at requests of high level Services Providers (SP) addressed to VCAN Providers (CANP) VCANs: Implemented as parallel logical data planes customised for different content types Content Awareness routing takes content-type or even name into account, not just location address 5
1. ALICANTE System Architecture (cont d) Environments: User (UE) : End-Users terminals Service (SE): Service and Content Providers Network (NE), CAN Providers and Network Providers Parallel logical Planes Actors: End-User (EU) Content Provider (CP) Service Provider (SP) Network Provider (NP) CAN Provider (CANP) Environment : groups of functions defined around the same functional goal
1. Alicante System Architecture (cont d) SrvMgr@SP Home Box + SP Environment Service Provider (SP) 1 CANMgr1 3 2.1 CANMgr2 2.2 CANMgr3 3 CAN layer Mgmt. CAN Provider (CANP) Intra-NRM@NP 3 Intra-NRM@NP Network Provider (NP) MANE 4 4 CR 4 5 HB2 EU host Content Server/ HB1 CND1 Multi-domain VCAN IntraNRM Intra-domain Network Resource Manager Intra-NRM@NP CND3 Media flows CND2 Access Network Media Aware Network Element Core Router
1. Alicante System Architecture (cont d) CAN Manager workflow 1. Request from SP to create mono-multi-domain VCAN to an initiator CAN Mgr 2. CAN Managers negotiate horizontally for resource provisioning TNRDP provides necessary information to support negotiations Then a combined algorithm is used by CANMgrs to map the VCANs onto real network topologies 3. Commands are sent vertically to each IntraNRM for installing VCANs configurations in routers 4. Each IntraNRM send vertical commands to install network policies (ingress/egress) to MANEs and also to core routers
2. Topology and Network Resources Discovery Protocol CAN Manager needs information for building VCAN Mappings Topological ( multi-domain graph) Resources (e.g bandwidth) Two possible modes of gathering information On-demand CAN Manager asks for info on each VCAN creation Proactive on timer/event, information is distributed over the network TNRDP proposal : proactive distribution of information between domain managers
2. TNRDP (cont d) Assumptions IntraNRM knows and transmits vertically information about internal paths (e.g based on pre-provisioned MPLS LSPs) Only CAN Mgrs participate in TNRDP TNRDP doesn t handle security/reliability CANMgr-s identities known statically, each one knows all its neighbors All domains are VCAN capable TNRDP Requirements Scalable - in terms of traffic overhead/number of CAN Managers Accessible - any CANMgr can find out information about any part of the network Parallel simultaneous changes in different parts of the network are handled Stability information should be consistently updated if changes appear in different domains
2. TNRDP (cont d) Design Each CANMgr communicates directly to its neighbors and generates Network State Advertisement (NSA) NSAs are generated periodically or triggered by other messages Each NSA received by a CANMgr is combined with its own information and transmitted further Hello is used for connection keep-alive TNRDP is stateful Each neighbor can be in either one of four states: Listen, Waiting for connect/disconnect confirm, Connected Any CANMgr may initiate Connect/Disconnect When a disconnect is received all info on links to that domain are erased If this domain is a leaf one, then all info on this domain is purged.
2. TNRDP (cont d) Messages Connect connection between two CANMgrs neighbors Disconnect disconnect to a neighbor NSA broadcast to all neighbors to update local graphs Confirmation confirmation of active messages Error signals syntax or semantic errors Hello keep-alive connection Message format: Type, Seq_no, Src_Mgr-Id, Dest_Mgr_Id, Data_length, Data Sequence numbers are increased in order by each sender
2. TNRDP (cont d) Example- message sequence chart CANMgr1 CANMgr2 CANMgr3 CANMgr4 Connected Connect Connected Conf NSA1 ( ) Connected NSA2 ( ) Update topology Update topology Conf Conf NSA2 ( ) Conf Update topology NSA3 ( ) Conf Update topology Graph: 1 <-> 2 <-> 3 <-> 4 Connections: 2-3, 3-4 1 wants to join and sends a connect to 2
2. TNRDP (cont d) Implementation and preliminary performance evaluation The protocol was implemented in Java for simulation and evaluation Communication complexity for a linear network O(D 2 ), D is the number of domains Communication complexity for a network of random topology O(D*n*d), D number of domains, d diameter of network in domains/hops, n number of connections for each domain
3. Conclusions A proactive Topology and Network Resources Discovery Protocol has been specified and designed and implemented To be used between domain managers in order to support creation of multi-domain Virtual Content Aware Networks Scalable solution This is a Software Defined Networking- oriented solution CAN Managers can be replaced by SDN Controllers Separation between the Data and Control Plane Partially centralized management Possible future development: migration towards full SDN approach Develop/use a Network Operating System Replace the vertical protocols controlling routers with OpenFlow
Thank you! Questions?